Rheological effects related to neo-fracturing processes in rock masses

(1)

Faculty of Mathematical, Physical and Natural Sciences

Department of Earth Sciences

Ph.D. in Earth Sciences

XXXII cycle

Rheological effects related to neo-fracturing processes in

rock masses

Ph.D. student

Danilo D’Angiò

Advisor

Prof. Salvatore Martino

Co-Advisor

Ph.D. Luca Lenti

December 2019

(2)

Abstract

Rockfall hazard is one of the main natural hazards in mountainous areas and along transportation routes. Roads and railways interruptions, as well as damages of buildings, are among the main inconveniences due to the detachment of unstable sectors of highly jointed rock masses. Rockfalls are the result of the combined action of the rock mass creep and of the natural and anthropic solicitations, which lead to the accumulation of inelastic strain within the rock mass and to the formation of new fractures or to the extension and movement of the pre-existing ones (rock mass damaging phenomena). The understanding of rock damaging processes through the microseismic monitoring of rock slopes predisposed for instabilities events can help in defining proper risk mitigation strategies. With the aim of assessing rock mass damaging phenomena, this PhD thesis proposes an analysis of the damping ratio associated with the microseismic emissions recorded in two test sites located in central Italy. Three monitoring campaigns have been conducted: two at the Acuto quarry test site, where the vibrational behaviour of a 12 m3_rock block partially detached from the back rock wall has been investigated; one at the Terni-Giuncano railway test site, where a rock mass close to a railway was studied to analyse the effects produced by the repeated trains transit. A STA/LTA event detection algorithm has been implemented for the recognition of the microseismic emissions from the seismic datasets acquired in continuous mode and with a sampling frequency of 2400 Hz. The damping ratio of the microseismic emissions, filtered in monofrequential waveforms, was evaluated; in the following, the daily mean damping values for each frequency were compared in respect to the environmental parameters monitored on site. No irreversible trend variations were observed, but significant variations related to transient processes were detected. It is deemed that the proposed approach can be applied on yearly seismic dataset and environments exposed to natural and anthropic forcing actions, to be furtherly tested and validated. Additional analyses were carried out on the train transit recordings. Each train passage was analysed in terms of: a) RMS value of the recording; b) cross-correlation between couples of sensors. The observation of the RMS and cross-correlation time series over time confirmed the unvaried long-term vibrational behaviour of the rock mass, according to the results of the damping analysis. In addition, the seismic noise was investigated by computing the average noise in one-minute intervals filtered at specific frequency bands. The cumulative of the filtered intervals allowed to determine the main energised frequency band, which resulted to be the one comprised between 500 and 1000 Hz.

(5)

5 The derivative of the averaged noise was compared with the environmental parameters recorded in the two test sites. A marked correlation between the variation of air and rock mass temperature and the derivative of the noise in the frequency band comprised between 0.5 and 30 Hz was noted, while the correlation is feeble or lost for the other frequency bands considered. The correlation observed between thermal cycles and ambient noise variations is in agreement with previous bibliographical studies; moreover, a differentiation in the vibrational response for the different frequency bands has been detected.

(6)

6

1 Introduction

Landslides are responsible of damages, casualties, traffic interruptions and economic losses all over the world, especially in mountainous areas (Guzzetti 2000, Nadim et al. 2006, Kirschbaum

et al. 2010, Kirschbaum et al. 2015, Froude & Petley 2018), where landslide hazard is one of the

major among the natural hazards. In fact, Froude & Petley (2018) report that almost 5000 non-seismically triggered fatal landslides occurred between 2004 and 2016, including several landslide typologies and triggering actions (Figure 1).

Figure 1: Top: number of non-seismically triggered fatal landslide events from 2004 to 2016 by country (top) and cumulative number of the recorded events (bottom) (Froude & Petley 2018).

In particular, hazard due to rockfalls, is the hardest one to predict among the several landslide typologies, because of the very few precursory phenomena and the short time available to react in respect to these rapid mass movements (Varnes 1978, Sättele et al. 2016) that can involve

(7)

7 from huge (slope scale) to very small (less than 1 m3_{) volumes of rock. Generally, most of the} rockfalls originate from the detachment of small portions of rock by an intensively fractured and jointed rock mass, because of the contribution of environmental (rainfalls, daily and seasonal thermal cycles, earthquakes) and anthropic (impacts, explosions, vehicular traffic) factors to which is added the role played by the long-term creep-driven deformations acting from slope to single joint scale. Considering the probability of small size rockfalls is higher than the one of greater mass movements, proper risk mitigation strategies should be adopted in the environments most sensitive to these kind of phenomena. In fact, especially in mountainous villages and along transportation routes, stabilization of slopes, the employment of retaining nets and protection barriers (Gottardi et al. 2011, de Miranda et al. 2015) are the most common remedies used to face the occurrence of rockfalls. In addition to these interventions, designed to reduce the vulnerability of infrastructures and of other human activities threatened by rockfalls occurrences, hazard mitigation is the subject currently managed by research activities not only in the field of earth sciences but in more different contexts, as those ones concerning tunnelling, quarrying, drilling and excavation activities.

Indeed, by taking into account the eventuality of injuring or killing people as well as the economic losses deriving from road or railway interruptions, the investigation of the processes leading to rock mass instability can provide useful contributions to the understanding and, subsequently, the prevention of rockfalls. These last, are generally triggered because of the accumulation of inelastic strain within the rock mass structure, which is responsible for the formation of new fractures or to the growth of the pre-existing ones. Such a process, known as rock mass damaging (RMD), leads to the deterioration of the mechanical parameters of the rock mass (as the friction angle and the elastic modulus), thus being at the base of the processes carrying to rockfalls. A contribution suitable for the understating of RMD effects is given by the analysis of stresses, displacements and passive seismicity characterizing the rock mass, which allow to derive indications on the stability conditions of the monitored object. Especially microseismic monitoring is an affirmed and widespread technique more and more adopted in several ambits, because of the capability of providing information on the modification of the internal structure of a medium before that it becomes evident to other monitoring devices. In fact, as sketched in Figure 2, the phenomenological evolution of the rock mass damaging processes can be

(8)

8 monitored through dedicated microseismic monitoring network, aimed at the understanding of the initial steps of them, which is one of the current topic of slope stability analysis.

Figure 2: Sketch of the processes of growth, expansion and opening of fractures within a rock mass, which progressively lead to rock blocks separation and detachment. These phases can be investigated with the deployment of a microseismic monitoring network.

In such a research framework, this PhD thesis reports the analysis of three microseismic monitoring datasets acquired in continuous mode on two rock masses, a first one located in an abandoned quarry in the Acuto municipality and a second one along a railway near to Terni, in central Italy. In particular, the focus of this thesis concerns the investigation of how the daily or periodic recurrence of perturbative events, both natural and anthropic, can affect the rock mass stability over time, as a consequence of the progressive accumulation of these solicitations, potentially leading to inelastic strain amassing. Aiming at assessing potential variations in the fractures network (i.e. damaging phenomena), a physical-based approach is presented, dedicated at evaluating the damping ratio associated with the microseismic emissions registered on the rock masses studied, for specific frequency of interest. The trend of the damping ratio over time is analysed and compared with meteorological and seismic data, besides, for the Terni test site the cross-correlation and the root mean square (RMS) values of the train transit records were derived as comparative and alternative tools for the presented methodology. Furthermore, a technique based on the data analysis of the seismic ambient noise recorded in both the test sites is presented and discussed. A comparison of the averaged seismic noise amplitude filtered in specific frequency bands with the environmental parameters registered on-site has been done, and their correlation investigated.

The proposed methodology, based on the analysis of the damping ratio, could be considered and applied on yearly-lasting microseismic dataset in environments subject to rapid

(9)

9 morfoevolutionary changes, as coastal cliffs and high mountain environments, to slopes already involved in landslide processes, as well as to rock masses subjected to recurrent man-induced vibrations, as for quarrying activities or to trenches in proximity of roads and railways.

(10)

10

2 Microseismic monitoring of rock masses: an overview

The investigations of the rock masses aimed at understanding the evolution of their deformational behaviour and the potentiality of generating rockfalls have increased more and more in the last years, also favoured by the development of novel inspections methodologies and by new technologies and devices. In fact, literature widely deals with the topic of rock masses monitoring, both for landslide risk reduction on natural slopes and for the safeness of quarry, mining and tunnelling activities that involve the excavation of rock masses.

In general, rock mass monitoring can be performed through several approaches, taking into account the extent of the area and the processes to be observed. Commonly, the evolution of the stress-strain conditions of the rock masses is investigated by means of a) extensimeters, strain-gauges, inclinometers that are at the base of classical monitoring apparatus (Loew et al. 2017); b) laser scanners, photogrammetry, terrestrial and satellite interferometry (Abellán et al. 2010, 2014, Barla et al. 2010, Strozzi et al. 2010, Fanti et al. 2013, Salvini et al. 2013, Gigli et al. 2014); c) accelerometers, geophones and other seismic devices to analyse the passive seismicity of the rock mass as well as the signals related to rockfalls occurring in the monitored area (Lacerda et al. 2004, Meric et al. 2005, Godio et al. 2006, Jongmans & Garambois 2007, Willenberg et al. 2008, Travelletti et al. 2013, Pazzi et al. 2019).

Especially geophysical techniques, being minimally invasive approaches able to characterise the evolution of the stress-strain processes acting within a medium, and thus suitable for providing additional information on it in respect to other approaches, are widely used for field monitoring activities. In fact, geophysical techniques can allow to detect variations in the physical parameters characterising the internal structure of the rock masses, before they become evident to a more classical displacements monitoring system or to laser scanners and other remote sensing ones.

It is clear that geophysical measurements depend on the complex interaction between the seismic waves and the studied medium, which basic principles are reported in paragraph 2.1. In the large field of geophysical methods, the development of seismic sensors with higher sensitivity, enhanced measurement accuracy and smaller dimensions, permitted the

(11)

11 investigation of the entire frequency spectrum covered by the motion waves, favouring the spreading of microseismic (MS) and acoustic emission (AE) techniques.

These methodologies consist of the detection, location and analysis of low energy seismic signals associated with the generation of new fractures and the movement and propagation of the existing ones that lead to the progressive failure of rock masses. Both the techniques allow to indirectly investigate the internal structure of the studied medium, mainly differing in the frequency range of the seismic motion recordable, which is higher for the AE signals in respect to the MS ones (Cai et al. 2007) (Figure 3).

In particular, the microseismic monitoring (MS) technique currently represents an affirmed diagnostic tool to detect vibrational signals in several contexts, as structural monitoring, mining excavation (Sun et al. 2012), tunnelling (Tang et al. 2010), and slope stability assessment as well (Papini et al. 2009, Colombero et al. 2018). The field of investigation of this practise is related to the signals characterised by a frequency content from 101_{to 10}3_{Hz, being comprised between} the classical ambit of earthquake and seismic noise studies, dealing with lower frequencies, and the acoustic emission domain, characterized by frequencies greater than 104_{Hz (Figure 3). Since} in this thesis is proposed an analysis of seismic signals up to a thousand Hz, the state of the art of the microseismic monitoring technique is presented in paragraph 2.3, focusing on the main works related to the monitoring of rock masses.

(12)

12 In addition to this technique, a recent development in the analysis of the seismic ambient noise measurements has been done. In fact, several studies deal with the information retrievable from the ambient noise and the correlation of its modification associated with changes in the characteristics of the medium, as thermal and hydrological variations (Larose et al. 2015) and also with the characterisation of rock slopes (Burjánek et al. 2010, Pilz et al. 2014, Kleinbrod et

al. 2019). A summary of the works concerning the analyses of rock masses through ambient noise

techniques is presented in paragraph 2.4.

2.1 Seismic wave propagation in rock masses

The propagation of surface waves (Rayleigh and Love) generated by any kind of source of vibrations, both natural (earthquakes, rockfalls) and anthropic (impacts and industrial machinery), strictly depends on the physical and geometrical characteristics of the crossed medium. In particular, the presence of interfaces and interruptions within the medium, as well as the transition among different media, are at the basis of energy dispersion phenomena. By taking into account the rock masses, the main characteristics that influence the wave propagation concern the rock lithology, the presence of several lithotypes characterised by a marked difference in the acoustic impedance and the existence of fractures and joints in the rock mass. Generally, when a seismic wave encounters a discontinuity is subjected to reflection and refraction phenomena. Such effects are responsible for energy dispersion and seismic wave attenuation or amplification. In particular, within a rock mass, because of the plasticity, the seismic wave is attenuated both by the geometric spreading and by the non-linear and viscous damping actions. Wu et al. (1998) studied the propagation characteristic of blast-induced shock waves by deploying several accelerometers along specific alignments on a rock surface, finding out that the aperture, the number of joints and the incident angle of seismic waves on them, are among the main factors controlling the wave propagation. Moreover, in agreement with Hao et

al. (2001), they highlight that the attenuation of the waves is higher in the direction

perpendicular to the strike of the joints than in the one parallel to them. Besides these basic principles, on which the approach proposed in this thesis is based, other specific studies have been carried out in the ambit of the mining research and of the wave attenuation for slope stability assessment. A review of the theoretical methods for the analysis of wave propagation

(13)

13 across jointed rock masses was presented by Perino et al. (2010). Li et al. (2010) performed laboratory testing and numerical modelling on the wave propagation across a filled rock joint observing that also an higher water content results in a lower transmission of the stress waves. A numerical analysis combined with tests conducted with a shaking table on a scaled model of a jointed rock slope was performed by Che et al. (2016), which measured the wave propagation at several positions of the reconstructed slope model. The Authors observed a delay in the propagation of the transmitted waves because of the presence of joints, emphasizing the concept that the number and spatial distribution of the joints are responsible for the detected time delay. This key notion has been taken into account in the development of the approach presented in paragraph 3.3.2.

2.2 Rock mass damage

The understanding of the processes that lead to the formation of new fractures and to the extension and movement of the pre-existing ones is fundamental for the assessment of the stability conditions of rock slopes and for the safeness of the engineered projects that insist on rock masses, as transportation routes, mines and dams. In fact, the accumulation of damage over time and its spatial localisation lead to a progressive degradation of the rock mass mechanical parameters, like strength and moduli, consequently causing an amass and redistribution of the stresses that can potentially evolve in failure processes. Within a jointed rock mass, the rock mass damaging processes act at different scales, causing the opening of the already-existing fractures (from millimetric to decametric extent) (Klein et al. 2008, Bretschneider et al. 2013) or the formation of new cracks (Kranz 1983). Generally, the stress applied on rocks, associated with a temporal factor, is the main parameter responsible for the fracturation phenomena. In addition to these creep-driven deformations, the rock mass damage can be rapidly increased because of natural and anthropic forcing actions that can act in a discontinuous or continuous way.

Among the factors responsible for the genesis of micro-fractures within the rock masses, a main role is played by earthquakes (Gischig et al. 2012, Koukouvelas et al. 2015, Martino et al. 2017, Romeo et al. 2017), blasting (Wei et al. 2009, Wang et al. 2018), tunnelling (Cai et al. 2004) and mining activities (Wang et al. 2015), rainfalls, daily thermal cycles and freeze-thaw ones (Matsuoka & Sakai 1999, Gunzburger et al. 2005, Mateos et al. 2012, Collins & Stock 2016, Dietze

(14)

14

et al. 2017) that are combined and superimposed with creep-driven deformations (Brückl &

Parotidis 2005, Grøneng et al. 2010, Xu et al. 2014).

Therefore, it is important to study both the time-dependent behaviour as well as the transient variations occurring within rock masses due to external solicitations (Fiorucci 2018). These factors jointly affect the rock mass from the slope scale (Deep-sited gravitational slope deformations - DSGDS) to the micrometric scale (fractures growth and formation), causing rock mass damage phenomena that cumulate over time, showing deformational effects at different time stints (Figure 4).

Figure 4: Sketch of the deformational processes acting on a rock mass at different scales and time stints. In addition to the rock mass creep, also external solicitations contribute to rock mass damage.

The stress-strain conditions responsible for the initiation of the fracturation phenomena have been widely investigated, with laboratory testing on rock specimens (Amitrano 2003, Hoek & Martin 2014, Cao et al. 2015), numerical modelling (Scholtès & Donzé 2012, Lisjak & Grasselli 2014) and on-site seismic surveys (Walter et al. 2012, Colombero et al. 2018).

(15)

15 In particular, the effects produced by thermal variations have been deeply investigated (Prick 2003, Yavuz et al. 2010, Lan et al. 2013, Akdag et al. 2018), because of the persistent action on the rock masses. Bakun-Mazor et al. (2013) studied the sliding mechanism of rock blocks partially detached from a rock slope, finding that the thermal factors may have been more important than seismic ones in controlling the displacements of the rock blocks.

The damaging phenomena are also studied at the specimens scale to observe and understand the loading conditions leading to the formation of the fractures. In the following, the mechanical properties of the rock are derived, scaled and inputted for numerical simulations.

Amitrano & Helmstetter (2006) realised a damage-based model to simulate the brittle creep of lab specimens by using a finite element model. The damaging of the elements constituting the model was reproduced by progressively decreasing their Young modulus, with the aim of simulating the effect of increasing crack density. Successively, this methodology was applied to simple slope geometries by Lacroix & Amitrano (2013) and to a real slope scale by Riva et al. (2016), which also considered the fracturation and varying damage properties of the medium. Plenty of studies deal with laboratory testing and numerical modelling (Eberhardt et al. 2004, Heap et al. 2009, Scholtès & Donzé 2012, Brantut et al. 2013, Xu et al. 2014, Yang et al. 2015), moreover an increasing number of works is focused on the direct investigation of damage phenomena at the rock mass scale through on-site microseismic surveys.

2.3 Fields of application of microseismic monitoring

As previously introduced, the microseismic monitoring has several fields of application, especially concerning the study of the rock masses involved in mining and tunnelling activities as well as characterised by slope instabilities. A MS event is a low-energy seismic signal associated with a plastic deformation occurring in a medium, because of the accumulation of elastic energy over a threshold depending on the properties of the material. The rapid release of the accumulated stress is associated with the formation of new fractures or the sliding and propagation of the existing ones. The energy radiated by the microcracks can be detected with proper seismic devices, as accelerometers and geophones, characterised by a wide resolution in frequency and high sensitivity and accuracy. The sensors, mainly three-component ones, are deployed in several

(16)

16 array configurations, in order to enclose (in underground excavations and mines) or to cover (in slope faces and cliffs) the portion of the rock mass to be monitored. Frequently, the environments where the microseismic monitoring is performed are characterised by harsh conditions, and several issues have to be faced: the placement of the sensors, a sufficient data storage capacity and the availability of a stable power supply to guarantee the continuous data recording. These characteristics are progressively harder to find when passing from partly anthropised environments, as mines, to strictly natural environments, as slope faces.

The processing of the recordings is firstly focused on the recognition of the signals referred to the rock mass fracturing, by manual or automatic identification based on the characteristics of the original waveform recorded. Among the automatic identification procedures, STA/LTA based algorithms are extensively used and diffused (Withers et al. 1998, Vaezi & Van Der Baan 2015, Li

et al. 2016) while machine learning techniques (Shang et al. 2017, Chen 2018, Dong et al. 2019,

Lin et al. 2019, Peng et al. 2019) are rapidly increasing and improving the microseismic events detection.

Once the MS events are recognised, they can be classified, located and the source parameters calculated. The classification of the MS events is mainly based on the shape, duration and energetic characteristic of the signal. The review and the analysis of multiple microseismic events recordings registered on slopes affected by different slope instabilities (slides, falls, topples and flows) led Provost et al. (2018) to propose a classification of the various typologies of microseismic signals.

By referring to data analysis and processing techniques, the counting of the MS events and their characterization in terms of energy content, spectral amplitude and other parameters derived from the seismological ambit, as the peak ground acceleration, are the features mainly taken into account. By monitoring the derived parameters over time, it is possible to assess the evolution of the stability conditions of the studied rock mass, permitting the definition of the hazard associated with the collapse of rock portions.

The mechanical properties of the rock mass, derived from the analyses carried out on microseismic data, can be also used as input for numerical simulations in which the parameters of the model are continuously updated with the real damage state of the rock mass. This approach has been recently applied by several Authors (Bozzano et al. 2013, Xu et al. 2014, Tang

(17)

17

et al. 2015, Zhao et al. 2017, 2019, Zhuang et al. 2019) in order to obtain a better estimation of

the failure time of rock slope portions.

Numerous studies have been carried out in the fields of quarrying and tunnelling activities (Wesseloo & Sweby 2008, Tang et al. 2010, Sun et al. 2012, Lu et al. 2013, Carlà et al. 2017, Zhao

et al. 2018, Peng et al. 2019) and for the management of construction projects as dams and

powerhouse caverns (Tang et al. 2010, Xu et al. 2015, Dai et al. 2016, Liu et al. 2018). In these cases, the microseismic monitoring helps in evaluating the progressive strain effects of rock masses affected by excavation activities as well as by the varying stress conditions exerted by a reservoir basin.

A vast bibliography also concerns the microseismic monitoring of unstable rock slopes (Spillmann

et al. 2007, Tang et al. 2010, Occhiena et al. 2014, Ma et al. 2017, Yuan-hui et al. 2018), which

aim is prevalently the investigation of the phenomena and environmental conditions responsible for the microseismicity. Recent studies were also focused on the microseismic monitoring of rock slopes close to railways (Xu et al. 2017, Yan et al. 2019), as rockfall hazard represent a threaten especially for the transportation routes located along cliffs and in man-excavated trenches (Budetta 2004, Ferlisi et al. 2012, Michoud et al. 2012, Macciotta et al. 2017, Messenzehl et al. 2017, Martino et al. 2019). In fact, the vibrations induced by trains may act as a repeated solicitation affecting the rock mass stability when long periods are considered.

Helmstetter & Garambois (2010) pointed out a correlation between the occurrence of rainfalls and the seismicity referred to rockfalls, highlighting a rapid response of the rock mass even with tiny amounts of precipitations. A similar correlation was found by Arosio et al. (2018), that analysed a three-year microseismic dataset acquired on an unstable rock slope in Northern Italy. Other studies are focused on the investigation of the dynamic behaviour of unstable rock slopes, by analysing the microseismic emissions as well as the seismic ambient noise recorded prior of the collapse of the investigated sectors (Got et al. 2010, Lévy et al. 2010, Levy et al. 2011, Bottelin, Jongmans, et al. 2013, Bottelin, Lévy, et al. 2013, Valentin et al. 2017). In particular, Got et al. (2010) and Levy et al. (2011) analysed the pre-failure behaviour of an unstable rock cliff, both recording an increasing microseismicity and observing significant variations of the seismic noise amplitude in the phases prior of the collapse.

(18)

18 Lots of these studies show how the variations detectable in the frequency spectrum, is a distinctive characteristic preceding collapse events of unstable rock blocks. Although the observed frequency shifting is an important feature to be considered for risk management strategies, this characteristic appears only in the final status of the deformation, when a significant mass of material is ready to be released. Moreover, in the perspective of previsional numerical simulation, the frequency data are not directly utilisable.

In order to investigate also minor instabilities phenomena, it is deemed that the analysis over time of the damping ratio associated with the microseismic emissions is suitable for obtaining information related to the modification of the internal structure of the rock mass. An increment of the damping value with time may indicate a tiny change into the micro-fracture network of the rock mass, which can precede instability phenomena.

2.4 Ambient noise measurements

In the last decade, several studies were addressed to the characterisation of the seismic ambient noise, especially for seismic hazard purposes. In fact, the ambient noise is a peculiar attribute that describes the vibrational behaviour of any medium, allowing to retrieve several information on the subsurface in different fields of application, from site seismic response to slope stability. The seismic ambient noise is the result of the different vibrations produced by several and disparate sources, both natural (sea waves, tides, wind) and anthropic (blasts, traffic, industrial machinery). Each source of vibration contributes to enrich the ground vibration with its own energy, carried by mechanical waves and associated with a specific frequency content.

Several studies deals with the categorisation of the sources constituting the seismic ambient noise (Bonnefoy-Claudet et al. 2006 and references therein). In their review, Bonnefoy-Claudet

et al. (2006) report the general agreement of the specific literature in defining the seismic noise

as prevalently constituted by surface waves. In particular, daily and weekly variations at frequencies higher than 1 Hz are related to human activities, while longer period changes associated with lower frequencies (between 0.005 and 0.3 Hz) are due to natural factors (oceanic waves and meteorology).

(19)

19 The composition of the seismic noise and the influences on its modifications were investigated because of the several ambits of application for which is used, among which the seismic microzonation is one of the most important (Nakamura 1989, Bard et al. 2010). Ambient noise has been also used for the investigation of landslides (Mainsant et al. 2012, Pilz et al. 2014, Collins & Stock 2016, Zare et al. 2017, Hussain et al. 2019) and unstable rock slopes (Burjánek et al. 2010, Moore et al. 2011, Panzera et al. 2012, Galea et al. 2014, Kleinbrod et al. 2019). The main outcomes consist of the reconstruction of the fracturation pattern of the monitored rock slope and the characterisation of landslide bodies or damaged zones in terms of geometrical and dynamic properties.

In addition to the spatial and dynamic information retrievable from ambient noise analysis, also a relationship with the variations occurring in the environmental parameters has been found. Larose et al. (2015) pointed out the correlation among the ambient noise and the thermal variations in the subsoil, buildings and even rock columns. A similar outcome was found by Colombero et al. (2018) that observed the correlation between daily and seasonal temperature fluctuations and the resonance frequency values of unstable portions of an highly-jointed rock mass located in Northern Italy.

In this thesis, the variations in seismic noise within specific frequency bands were compared in respect to the environmental parameters recorded in the test sites, to check the existence of a correlation for all the considered frequencies.

(20)

20

3 Materials and methods

This PhD thesis deals with the analysis of continuous microseismic datasets recorded on rock masses, being focused both on the investigation of the physical characteristics of the microseismic events and on long-period ambient noise variations in specific frequency bands, aiming at recognizing rock mass damaging phenomena. In particular, a novel physically-based technique focused on the investigation of the damping ratio associated with the microseismic emissions is proposed (Damping Analysis, paragraph 3.3.2). The study of ambient noise variations deepen an already proposed data analysis technique (Frequency Band Ambient Noise Disaggregation Analysis, paragraph 3.3.4).

To study the RMD processes acting on rock masses, it is needed to collect wide and multi-sensors datasets in natural test sites, lasting from months to years preferably, in order to take in consideration all of the different variables and a sufficient time span necessary to observe potential modification in the monitored deformational/vibrational parameters. Since the processes of fractures formation and propagation can act very slowly, even years-lasting datasets could show no significant results in the monitored parameters, thus the selection of the case study for testing a novel technique has to be accurately done. In fact, contrary to laboratory experiments performed on rock specimens to understand fractures formation conditions (Amitrano 2003), since in natural settings it is not possible to accelerate the physical processes acting within the medium, unless with experimental activities aimed at forcing some boundary conditions (Fiorucci 2018), it is crucial to choose fast-evolving environments. Even if natural settings are less controlled under the point of view of the parameters and boundary conditions that characterise the case studied, in situ microseismic monitoring provides the opportunity, and the challenge, of taking into account the variability of the contribution of environmental and anthropic factors acting at different rock mass scales.

To test the reliability of the proposed technique and in order to obtain a wide collection of microseismic events, specific microseismic monitoring campaigns were performed on the rock masses cropping out in two natural test sites located in central Italy. The first one is the Acuto test site, situated in an abandoned quarry in the municipality of Acuto, about 100 km SE far from Rome; the second one is located along the Terni-Giuncano railway, close to the Terni town, 150 km NE far from Rome. The test sites (paragraph 3.2) were selected to be representative of two

(21)

21 opposite conditions: the Acuto test-site is mainly characterized by natural vibrations, as the abandoned quarry is far from roads and human activities; the Terni test-site was selected because very close to an active railway line, in which the contribution of the repeated train transits in soliciting the rock mass is added to the damaging effect due to natural actions. For the Terni-Giuncano railway test site, given the presence of trains transiting close to the rock mass, additional analyses have been done on the trains passage recordings (Cross-correlation and RMS analysis, paragraph 3.3.3), in order to investigate even the trend of these parameters over the monitored period.

Furthermore, in both the test sites rock samples were taken and shaped to obtain prismatic test pieces analysed by means of an ultrasonic testing device, to derive the P-waves propagation velocity (ASTM 2005) of the rock masses analysed.

In Table 1 are summarised the type of analysis performed in the two test sites; the microseismic equipment, the test sites and the methodology employed for the analysis of the microseismic dataset are introduced in the following paragraphs.

Table 1: Summary of the analysis performed in the two test sites.

Acuto test site Terni test site

Type/Target of the analysis Ambient noise MS Events Ambient noise MS Events Train passages Damping Analysis X X F-BAND analysis X X RMS Analysis X Cross-correlation analysis X PGA, PGV, Arias, FE X X X

3.1 Microseismic monitoring devices

The microseismic monitoring in the two test sites has been carried out through high sensitivity Bruel & Kjaer (BKSV) one-component piezoelectric accelerometers (type 8344) (Figure 5), cable connected with a signal amplifier SomatXR MX1601B-R coupled with a digital acquisition system

(22)

22 SomatXR CX23-R both produced by HBM (Figure 6). The connection between the signal amplifier and the accelerometers was done with twenty-meter long low noise cables. The main features of the sensors are reported in Table 2, while the dimensions and a calibration chart of a sensor are shown in Figure 5. From the latter, it is possible to remark the flat frequency response in the range 5-2000 Hz and the sensitivity of the sensor of 2500 mV/g, characteristics that allow to detect very low signals emitted by the monitored object. Besides, the very small dimensions of the accelerometers and the sealed structure made them proper to be used even in harsh environments.

Figure 5: Dimensions, characteristics and calibration chart of the BKSV 8344 accelerometer.

Even the digital acquisition system (Figure 6) has a robust and watertight design and it can record up until 16 different channel contemporaneously, with a sampling frequency up to 20 kHz and an internal storage capacity of 64 GB. Data are saved in a compressed SIE file format: an automatic saving is done every 5 seconds, avoiding to lose the recorded data in the case of power supply interruptions, furthermore, the data logger has the capability of automatically restart the acquisition with the set parameters when the electricity supply is restored. Moreover, in the

(23)

23 occasion of continuous monitoring and of very high sample rates, to avoid troubles in managing just one huge data file, it is possible to set the creation of several files according to a desired duration or to a specified time of the day. The system is also predisposed for the automatic sending of data packages through LAN connectivity, but both the test-sites are not reached by the internet, thus data download was manually done. In fact, as for all the microseismic monitoring campaigns conducted in natural environments, it was required to face some technical issues related to the absence of electricity and internet connection as well as the installation and maintenance of the microseismic equipment. For both the test sites, the time recurrence of the maintenance of the equipment and the data download was established in about two interventions per month.

Figure 6: Digital acquisition system SomatXR CX23-R (left) and signal amplifier SomatXR MX1601B-R (right).

In both the test sites, as detailed described in paragraphs 3.2.1 and 3.2.2, the sensors were deployed on the rock mass surface. Once that the spots to be monitored were individuated, 7 mm diameter holes were drilled perpendicular to the rock surface to host metallic 5 mm diameter studs, subsequently fixed by using epoxy resin. In the following, the accelerometers were tightened to the metallic studs, which guaranteed the coupling between the sensors and the rock surface.

Since both the localities are not reached by electricity, the power supply was guaranteed by means of a 105 W solar panel connected to an automatic voltage regulator and to a 40 Ah lead backup battery. Unfortunately, for reasons related to the exposition of the solar panel, the season, and the alternation of sunny, cloudy and rainy phases, some interruptions during the monitoring periods were unavoidable. The data logger, the signal amplifier, the backup battery

(24)

24 and the automatic voltage regulator were placed into a watertight box placed in proximity of the monitored rock masses.

In addition to the microseismic monitoring devices, also environmental data were collected by means of on-site installed weather stations, providing information about the external conditions affecting the monitored slopes.

Before introducing the methodology used for the analysis of the microseismic datasets, the two test sites will be presented in the next paragraph, to better specify differences and similarities of the two case studied and to show the configurations adopted with the sensors.

(25)

25 Table 2: Main dynamic, environmental and physical characteristics of the employed accelerometers.

Dy n am ic Ch ar act er is ti cs

Voltage Sensitivity (@ 159.2 Hz and 4

mA supply current) mV/ms–2 (mV/g) 250 ± 20% (2500± 20%) Measuring Range ms –2_{peak (g} peak) ±26 (2.6)

Frequency Range (±10% limit)

Amplitude Response Hz 0.2– 3000

Frequency Response Individual Frequency Response

on calibration chart

Mounted Resonance Frequency kHz >10

Transverse Sensitivity (@ 30 Hz, 100

ms–2₎ %

< 5 of the sensitivity of the axis in question

Transverse Resonance Frequency kHz 3.5

En vir o n m e n tal C h ar act e ri st ics

Operating Temperature Range °C (°F) – 50 to + 100 (– 58 to + 212)

Temperature Coefficient of Sensitivity %/°C + 0.05

Temperature Transient Sensitivity (3 Hz

LLF, 20 dB/decade) ms–2/°C 0.001

Base Strain Sensitivity (at 250 µ in base plane) Equiv. ms–2/µ (g/µ) 0.002 (0.0002) Magnetic Sensitivity (50 Hz, 0.038 T) ms–2/T (g/T) 0.5 (0.05)

Max. Non-destructive Shock ms

–2_{peak (g} peak) 3500 (350) Humidity 100% RH non-condensing Phy si cal Ch ar act er ist

ics _{Case Material} _{Stainless steel AISI 316– L}

Sensing Element Piezoelectric, Type PZ 27

Sealing Hermetically sealed

Weight (excluding cable) gram (oz.) 176 (6.2)

3.2 Case studies

The case studies are both located in the central sector of the Apennine chain (Figure 7) and were chosen in order to be representative of two opposite conditions: the first one implies a rock mass

(26)

26 and a prone-to-failure rock block mainly solicited by natural actions (Acuto quarry test site, also abbreviated as “Acuto” in the text); the second one is mainly devoted to the understanding of the contribution of anthropic vibrations given to a rock mass close to a railway line (Terni-Giuncano railway test site, also abbreviated as “Terni” in the text). For both the sites, the Research Centre for Geological Risks (CERI) of Sapienza University of Rome carried out different research activities devoted to the definition of rockfall hazard prevention strategies. In particular, the Acuto test site is directly managed by the CERI since 2015, while the Terni-Giuncano railway site has been individuated in agreement with the Italian National rail infrastructure manager (RFI), which allowed the access to the rail network for performing the experimentation planned for this PhD thesis. In both the test sites, preliminary investigations were carried out in order to choose and then predispose the sectors to be monitored.

Figure 7: Satellite view and location of the two test sites in central Italy.

A total of three microseismic monitoring campaigns, resumed in Table 3 and Figure 8 and extensively presented in the following paragraphs, have been carried out in the two test sites. In all the cases, the data acquisition was set in continuous mode, with a sampling frequency of 2400 Hz and for saving files every three hours; vibrational data were saved in g (m/s2_{) unit and in} Coordinated Universal Time (UTC or GMT, Greenwich Mean Time).

(27)

27 Table 3: Resume of the microseismic monitoring campaigns and of their main features. For the Acuto 2019 campaign, the duration of 145 days is referred to the dataset analysed until April 22nd_2019.

Campaign Beginning End Duration (days) Deployed sensors Measured component Sampling Frequency (Hz) Location of the sensors On-site weather monitoring Acuto 2018 23/02/2018 31/05/2018 97 6 NS 2400 Rock block Rock mass All the period

Terni 2018 04/06/2018 14/11/2018 163 3 - 6 EW 2400 Rock mass From

08/08/2018 Acuto 2019 28/11/2018 ongoing 145* 6 - 7 NS - EW - UP 2400 Rock block Rock mass All the period

Figure 8: Time spans of the microseismic monitoring campaigns and sensors deployed.

3.2.1 The Acuto quarry test site

In the ambit of the research activities conducted by the CERI, the abandoned quarry of Acuto (Frosinone, central Italy) was selected since Summer 2015 as test site for the installation of a permanent multi-sensor monitoring system on a prone to failure rock block, aiming at investigating the long-term rock mass deformations due to temperature, wind and rainfalls (Fantini et al. 2016, Fantini et al. 2017). The quarry is located in the municipality of Acuto (Figure

(28)

28 9), 100 km SE far from Rome, on a carbonatic hill at an altitude of about 750 m a.s.l. The site location corresponds to the carbonatic Ernici Mounts ridge, which is part of the Latium-Abruzzi platform sedimentary domain, mainly constituted by thick Meso-Cenozoic carbonatic successions. In particular, Mesozoic wackestones with rudists outcrop on the sub-vertical quarry front (Accordi et al. 1986), which is extended about 500 m with a height ranging from 15 up to 50 m.

The quarry wall is aligned along the E-W direction, while the northwestern sector of it is oriented at N 10. The geomechanical characterisation of the rock mass led to the identification of four joint sets (Fantini et al. 2016), here indicated according to dip direction/dip convention: S0 (130/13) corresponding to the limestone strata, S1 (270/74), S2 (355/62) and S3 (190/64). In the NW portion of the quarry, a 12 m3 _{intensely jointed protruding block exposed to SE has been} individuated as target for the installation of the permanent multi-sensor monitoring system in 2015 and as focus of the temporary microseismic monitoring performed in the ambit of this PhD thesis. The rock block is separated on its western portion (also named block side) by a main open fracture (oriented 115/90) from the quarry wall, while on the eastern part, even if limited by a secondary fault line, the block is more contiguous with the rock mass wall. For this reason, the front and the lateral side of the rock block resulted to be the more interesting sections to be monitored, first under the deformational point of view, then under the vibrational one.

The permanent multi-sensor monitoring system consists in:

 1 thermometer for the rock mass temperature, installed on the frontal face of the rock block;

 6 strain-gauges installed on micro-fractures of the rock mass;  4 extensimeters installed on open fractures;

 2 weather stations, installed at the foot and the top of the slope wall, equipped with an air thermometer, a hygrometer, a pluviometer and an anemometer for wind speed and direction.

(29)

29 Figure 9: View of the Acuto quarry wall (a) with the rock block selected as the focus of the multi-sensor monitoring system (red rectangle). Geological sketch (b) and satellite view (c) of the Acuto municipality and of the quarry area (blue rectangles).

All these sensors are connected to a data logger CR1000 Campbell Scientific, equipped with 24 acquisition channels and set to sample data every minute in the local time unit. Besides, a GPRS wireless connection system allows the automatic transmission of data packages to a local server every 4 hours and enables the remote control of the datasets. The deformational and meteorological datasets are acquired in continuous mode since Autumn 2015.

Several studies and investigations have been carried out so far, both focused on the predisposing conditions for rockfall release and the testing of new technologies and methodologies. Among them, the thermal response of the rock mass was investigated via a monthly infrared

(30)

30 thermographic survey (Fiorucci et al. 2018), in order to understand the potentiality of thermal cycles in releasing rock portions. Furthermore, aiming at recreating real rockfall hazard scenarios impacting on infrastructures, also a railway track was posed to test a prototypal optical device for the detection of rockfalls, based on a change detection algorithm working in a selected region of interest (i.e. the railway track and its surroundings) (Fantini et al. 2017). Besides, the reliability of the nanoseismic monitoring technique (Joswig 2008) in detecting rockfalls was also investigated (Hakes et al. 2018).

To carry on the investigations on rockfalls predisposing phenomena at the Acuto quarry, the 12 m3_{rock block was chosen as the focus of the microseismic monitoring for this PhD thesis.} Moreover, considering that the rock block is partially separated by the rock mass, it was deemed potentially subject to greater mobility in respect to the rock mass, and thus worthy to be investigated. Furthermore, the location of the quarry, far from anthropic disturbances and with the main quarry front exposed to south direction, makes this site ideal to study the evolution of deformation and fracturation processes mainly attributable to thermal cycles and natural actions.

To characterise the P-waves velocity of the rock mass, some rock blocks were sampled and shaped in 6 parallelepipedal test pieces. The ultrasonic testing executed following ASTM (2005) resulted in P-waves velocity ranging from 3923 to 4460 m/s resulting in an average of 4193 m/s. Two microseismic monitoring campaigns have been carried out: the first one took place from February 23th_{2018 to May 31}st_{2018 (Acuto 2018), lasting a little more than three months; the} second one started on November 28th_{2018 and is currently ongoing (Acuto 2019). A summary of} the placement adopted for the accelerometers for both the monitoring campaigns is reported in Table 4.

(31)

31 Table 4: Summary of the accelerometer deployments for the two monitoring campaigns conducted at Acuto test site. Full period is intended with the exception of the interruptions due to no power supply.

Campaign # Sensors on rock block

# Sensors on rock mass

Measured

component Monitored period

# Sensors in same position

Acuto 2018 1, 2, 3, 4 5, 6 NS Full period 6, 2

Acuto 2019 4 1 NS Full period 1, 4 5 2 EW Full period 6 3 UP Full period 7 - NS Starting 09/01/2019 Several interruptions

Figure 10: In b) is shown the rock block with the railway track posed at its base. The squared areas zoomed in a), c) and d) show respectively: the block side with a main open fracture on which are installed two extensimeters (pointed by the arrows); the data logger and the weather station located at the top of the slope; the block front on which is circled the position of the thermometer for the rock mass temperature, while the green arrows point to two strain gauges.

(32)

32 For the first monitoring campaign (Acuto 2018), six BKSV one-component accelerometers were deployed along an horizontal alignment with a spacing of about 15 cm: in this configuration 3 sensors were placed on the rock block (ID: 1, 2, 3), 2 were located on the rock mass wall (ID: 5, 6) and the last one was positioned at the passage between the rock block and the rock mass wall (ID: 4). In the adopted configuration, all the sensors were measuring the N-S direction (Figure 11). During the three-month monitoring period, it was recorded also an intense and unconventional meteorological event in the last days of February, in which a strong decrease of the temperatures was recorded.

Figure 11: Positioning and ID of the accelerometers for the first microseismic monitoring campaign (Acuto 2018) (a). The location of the sensors on the rock block side is also shown in the circled area in b). The digital acquisition system and the backup battery were put into a watertight box placed on the top of the slope wall (c).

For the second monitoring campaign (Acuto 2019) a different configuration of the accelerometers has been adopted, keeping only two sensors in the same position of the previous monitoring period (in particular sensors with ID: 6 and 2, that in this new campaign are respectively ID: 1 and 4) (Figure 12). This new configuration was chosen to point out the 3D vibrational behaviour of both the rock mass and the rock wall, by measuring along three

(33)

33 orthogonal components. The microseismic monitoring started by deploying six accelerometers following the scheme shown in Figure 12, in particular sensors with ID 4, 5, 6 were placed on the rock block, measuring respectively the N-S, E-W, and vertical components of the motion; the same components were measured by the sensors positioned on the rock mass wall (ID: 1, 2, 3). A new sensor (ID: 7) has been added on the eastern side of the rock bock aligned along the N-S direction. Due to technical issues, the new sensor was not available for all the duration of the monitoring period. For both the monitoring campaigns, the acquisition was set in continuous mode and with a sampling frequency of 2400 Hz. The maintenance of the equipment and the download of the data took place about twice per month.

Figure 12: Positioning and ID of the accelerometers for the second microseismic monitoring campaign (Acuto 2019) (a). The position of sensors 1 and 4 coincides with that one of accelerometers 6 and 2 in the previous monitoring campaign. A zoom of the block side and of the rock mass is shown in b), while the block front is reported in c).

(34)

34

3.2.2 The Terni-Giuncano railway test site

The national railway line, which connects Terni to Giuncano, is characterised by a single-track line passing through a narrow and deep-incised valley in a hilly area of Umbria region, in central Italy. This section of the railway is part of the Rome-Ancona railway, an important and historical route crossing the Apennines and linking the Tyrrhenian Sea to the Adriatic Sea. In fact, this railway is classified as “fundamental railway line”, because it allows to connect central Italy from east to west, both in terms of passengers and of goods. In particular, the railway comprised between the cities of Terni and Giuncano is affected by a number of transits per day of about 42 trains on weekdays and 25 trains on bank holidays, with a maximum velocity limited at 80-100 km/h. As usual, the majority of the transits is clustered during rush hours in the morning and in the afternoon, while the traffic is interrupted during nights. Because of the importance of this transportation corridor and considering that of the total length of 292 km, 121 km are on a single track, several parts of the line have been doubled in recent years, others are currently in progress and others are planned for the future. Among the latter, the development of the track comprised between the cities of Terni and Spoleto is intended with the realization of a 22.4 km long tunnel that would allow in speeding up the circulation, increment the railway traffic and avoid the path crossing along the flanks of the slopes. In fact, this sector of the Apennines is constantly affected by slope instabilities threatening the infrastructures (Cardinali et al. 2002, Guzzetti et al. 2003, Guzzetti et al. 2004, Guzzetti et al. 2004, Galli & Guzzetti 2007), because of the presence of highly jointed rock masses outcropping along the man-cut trenches.

Under a geological point of view, the considered area is characterised by folded and jointed rock masses involved in thrusting and faulting events originated by the Apennine chain genesis. In particular, the Terni-Giuncano railway runs along the Serra River valley in correspondence to a canyon cut by the homonymous river in the southern sector of the Martani Mounts, NE of the Terni geological basin. In this area, main tectonic elements as the Monte Torricella thrust and the Battiferro fault (Calamita & Pierantoni 1994, Bruni et al. 1995) constantly modified the stress-strain status of the rock formations that constitute the current slopes. The Scaglia Rossa formation, a Mesozoic limestone, outcrops along extensive sections of the railway trench, and even in the location individuated as test-site (Figure 13).

(35)

35 Field inspections and geomechanical surveys were carried out along the railway line between January and March 2017, to select the rock mass wall more suitable for installing seismic devices (Figure 14). Finally, the test-site was selected close to a tunnel, where no retaining nets were deployed to equip the rock walls and a visible block-size slope debris testified previous small-size rockfalls. The test-site is located at about 280 m a.s.l., where the railway line, running approximately along the N-S direction, passes 4.5 m far from a rock mass partially excavated to allow the realization of the northern entrance of the tunnel.

The highly weathered conditions of the outcropping rock mass resulted by the concurrence of the main joint patterns, also related to the outcropping parasitic folds, hydraulic conditions of joints and the presence of vegetation on the rock wall favoured by its north-facing. Despite the very complex fractures network and the alteration conditions, a geomechanical characterisation, performed following ISRM (1978), provided a number of joint per cubic meter (Jv) of 7-10 and Ib: 0.3-0.5, indicating the small to medium size of the blocks potentially releasable from the rock mass. Two main joint sets were individuated, which are respectively indicated according to the dip direction/dip convention: J1 (170/80) and J2 (350/78), both presenting an opening of 2-3 mm and with a terrain filling. For both the joint sets, a mean Joint Roughness Coefficient (JRC) of 8 was evaluated, while the Joint Coefficient Strength (JCS), resulted in an average interval of 44-46 MPa, according to Barton & Choubey (1977).

Furthermore, in order to derive the physical and mechanical properties of the outcropping limestone some of the rock block at the base of the slope wall were sampled to perform laboratory tests on them. In particular, some rock blocks were shaped in parallelepipedal 3x3x5 cm test pieces then used to derive the P-wave velocity; some others were employed for the assessment of the uniaxial compressive strength by means of the point load test. According to the hydrostatic weighing method (ISRM 1979), the test pieces were used for deriving the average density (ρ) of the rock, that resulted to be of 2630 kg/m3_{, corresponding to a natural specific} gravity (γn) of 25.79 kN/m3. The specific gravity of soil solids, determined following ASTM (2006), resulted to be of 26.62 kN/m3 _.

(36)

36 Figure 13: National Railway infrastructure in the Umbria region (a). A satellite view of the study area (green rectangle) is zoomed in b), where the railway line is highlighted in orange and also the location of the test site is marked. In c) the geological map of the area reported in b) is represented; the legend corresponds to: 1) Red earth and debris (Holocene); 2) River floods (Holocene); 3) Debris and alluvial fan (Pleistocene); 4) Sands and clayey sands (Plio-Pleistocene); 5) Clays and arenaceous clays (Marnoso-Arenacea Formation, Miocene); 6) Marly limestones (Bisciaro Formation, Miocene); 7) Marls and marly limestones (Scaglia Cinerea Formation, Miocene); 8) Marly limestones (Scaglia Rossa Formation, Eocene); 9) Marls (Marne a Fucoidi Formation, Cretaceous); 10) Limestones (Maiolica Formation, Cretaceous); 11) Limestones and marly limestones (Jurassic); 12) Limestones (Corniola Formation, Jurassic); 13) Limestones (Calcare Massiccio Formation, Jurassic). In d) the rock mass selected for the microseismic monitoring is shown.

(37)

37 Figure 14: Views of the inspected section of the Terni-Giuncano railway line. Many sectors of the railway pass close to rock masses equipped with retaining nets (b) or in deep excavated trenches (c). In a) a zoom of the material accumulated in the retaining net reported in b) is shown. During the maintenance interventions, small instability events occurred in the proximity of the railway line were also detected (d and e).

The ultrasonic test performed on 5 test pieces according to ASTM (2005), provided P-waves velocity ranging from 3580 to 4763 m/s resulting in an average of 4222 m/s. Moreover, the results obtained from the ultrasonic test were used for the determination of the porosities of the test pieces following the empirical relation by Fourmaintraux (1975), resulting in a total porosity comprised between 2.1 and 3.8%, with an average of 2.8%. The point load test performed on 8 rock samples following ASTM (2008) resulted in an average Uniaxial Compressive Strength (UCS)

(38)

38 of 65.32 MPa. All the parameters derived from laboratory and on-site testing are resumed in Figure 15.

Rock mass matrix

Natural specific gravity γn Specific gravity of soil solids γs Uniaxial Compressive Strength σc Average porosity n Average Vp kN/m3 _kN/m3 _MPa _% _m/s 25.79 26.62 65.32 2.8 4222

Rock mass joints

Joint set

Dip direction

and dip

Alteration

status Hydraulic conditions Opening Joint Filling JCS JRC

- ° - - cm - MPa -

J1 170/80 High Yes 0.3 terrain 46 8

J2 350/78 High Yes 0.2 terrain 44 8

Figure 15: Main geomechanical parameters derived from on-site and laboratory testing of the outcropping rock mass at the Terni test site and stereographic equal-angle projection (lower hemisphere) of the joint sets surveyed.

Following the obtainment of the authorisation for access to the railway line from the RFI and the predisposition of the rock wall, the microseismic monitoring started on June 4th_{2018. Initially, it} was planned the deployment of six BKSV accelerometers with a spacing of about 15 cm along a vertical alignment covering about 2 m, with the aim of studying the wave propagation through the rock mass (Figure 16). Unfortunately, during the installation phase, three sensors were not working properly, that is why the configuration was modified keeping only the accelerometers with IDs 1, 5 and 6, respectively the closer one to the ground and the uppermost ones along the alignment. In this way, the possibility of studying the delays in the wavepaths through the entire section of the rock mass monitored was maintained before the original six devices configuration was restored on August 3rd_{2018. All the accelerometers were measuring the vibrational} component oriented about E-W, which is the direction perpendicular to the slope face orientation.

In addition, on August 8th_{2018, a Davis Vantage Pro 2 weather station equipped with} pluviometer, air thermometer, and anemometer for wind speed and direction, was installed on-site to better characterize the local environmental conditions. The acquisition was set with a

(39)

39 sampling rate every 15 minutes, which allowed to do not saturate the internal storage memory before the scheduled maintenance and data download operations. A general interruption in the acquisition of the equipment was planned between September 4th_{2018 and September 18}th 2018, for the maintenance of the installed devices. A summary of the data collected during the monitoring period is reported in Table 5.

The microseismic monitoring system recorded until November 14th_{2018, when all the installed} devices were removed to be furtherly employed in the Acuto test site. Also in this campaign was adopted a sampling frequency of 2400 Hz and a power supply guaranteed by a solar panel connected to a backup battery. Unfortunately, this led to several interruptions in the last period of the monitoring, coinciding with the shortening of the days in the autumnal season jointed with the beginning of the most rainy period and the unfavourable northern exposition of the solar panel.

Table 5: Summary of the data collected during the monitoring campaigns.

Peri o d _From _{04/06/2018 04/08/2018 08/08/2018 04/09/2018 18/09/2019} To 03/08/2018 08/08/2018 03/09/2018 17/09/2019 14/11/2018 Da ta Microseismic 3 sensors: ID 1, 5, 6 6 sensors 6 sensors Acquisition interruption 6 sensors Weather Terni weather station Terni weather station On-site weather station Acquisition interruption On-site weather station

Air temperature and rainfalls intensity (cumulative mm) data related to the period previous of the installation of the on-site weather station and during the acquisition interruption were recovered by the hydrographical service of the Umbria region, which manages a weather station located in the Terni town, about 7 km far from the test site. Just one data per day was available but it permitted to fill the uncomplete weather dataset for the first two months of the monitoring.

(40)

40 Figure 16: Positioning and ID of the accelerometers employed for the microseismic monitoring campaign (b). A zoom on an accelerometer is shown in c) and the on-site weather station is presented in a).

In order to support the following data analyses, the timetables of the trains effectively transited on the railway line in the monitored period were requested to the RFI. Such documents were provided at the end of the monitoring period, reporting all the trains transited, their typology (passenger, freight, single locomotive), their identification number, and the time of effective transit at the Terni or Giuncano railway stations, between which is comprised the monitored rock mass. By joining all of the provided information, a database of all the transits was created, including not only the already-listed features of the trains, but also by assigning an identification number related to the train type and two progressive numbers referred to the total count of trains transited for each day and from the beginning of the monitoring. Furthermore, knowing the distance of the test site from the two stations, it has been possible to evaluate for each train the approximate time of transit near to the rock mass monitored, which helped in the identification of the trains in the seismic records. A few number of trains (37) has been removed from the database because of the incompleteness of the information in the provided timetables,